Graph-based RAG Enhancement via Global Query Disambiguation and Dependency-Aware Reranking

2025-06-07

AI Summary

Existing graph-augmented RAG methods rely solely on entity extraction for knowledge graph retrieval, leading to semantic omissions, relational misjudgments, and amplified hallucinations—thereby compromising response fidelity. To address these issues, we propose a Hierarchical Global Query Parsing and Dependency-Aware Re-ranking framework. First, we introduce fine-grained query decomposition via multi-level parallel-sequential joint path modeling. Second, we design the first explicit subproblem dependency-aware graph retrieval re-ranker. Third, we construct an end-to-end architecture integrating hierarchical query parsing, knowledge graph retrieval, dependency graph modeling, and LLM-coordinated reasoning. Evaluated across multiple RAG benchmarks, our method significantly outperforms state-of-the-art approaches, achieving substantial improvements in answer accuracy and factual consistency. Comprehensive ablations and cross-dataset experiments validate its generalizability and robustness.

Technology Category

• Data Mining & Knowledge Management: Intelligent Query Processing
• Natural Language Processing: Question Answering
• Knowledge Representation and Reasoning: Computational Complexity of Reasoning

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Graph Algorithms and Modeling for the Web: Querying, indexing, and retrieval in Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Contemporary graph-based retrieval-augmented generation (RAG) methods typically begin by extracting entities from user queries and then leverage pre-constructed knowledge graphs to retrieve related relationships and metadata. However, this pipeline's exclusive reliance on entity-level extraction can lead to the misinterpretation or omission of latent yet critical information and relations. As a result, retrieved content may be irrelevant or contradictory, and essential knowledge may be excluded, exacerbating hallucination risks and degrading the fidelity of generated responses. To address these limitations, we introduce PankRAG, a framework that combines a globally aware, hierarchical query-resolution strategy with a novel dependency-aware reranking mechanism. PankRAG first constructs a multi-level resolution path that captures both parallel and sequential interdependencies within a query, guiding large language models (LLMs) through structured reasoning. It then applies its dependency-aware reranker to exploit the dependency structure among resolved sub-questions, enriching and validating retrieval results for subsequent sub-questions. Empirical evaluations demonstrate that PankRAG consistently outperforms state-of-the-art approaches across multiple benchmarks, underscoring its robustness and generalizability.

Problem

• Misinterpretation of latent query information in graph RAG
• Irrelevant retrieval due to entity-level extraction limitations
• Hallucination risks from incomplete knowledge graph utilization

Innovation

• Hierarchical query-resolution strategy for global awareness
• Dependency-aware reranking mechanism for structured reasoning
• Multi-level resolution path capturing query interdependencies